Morphological divergence in fragmented populations: No evidence of fitness benefit

Raw code available at SW_Morphology

Study Design

Setup

Load all packages

# libraries
library(car)
library(coxme)
library(dplyr)
library(ehahelper)
library(ggeffects)
library(ggplot2)
library(ggpubr)
library(ggthemes)
library(lme4)
library(lmerTest)
library(readr)
library(sjmisc)
library(survival)

sessionInfo()

sessionInfo()

R version 4.3.0 (2023-04-21)
Platform: aarch64-apple-darwin20 (64-bit)
Running under: macOS Ventura 13.4

Matrix products: default
BLAS:   /System/Library/Frameworks/Accelerate.framework/Versions/A/Frameworks/vecLib.framework/Versions/A/libBLAS.dylib 
LAPACK: /Library/Frameworks/R.framework/Versions/4.3-arm64/Resources/lib/libRlapack.dylib;  LAPACK version 3.11.0

locale:
[1] en_US.UTF-8/en_US.UTF-8/en_US.UTF-8/C/en_US.UTF-8/en_US.UTF-8

time zone: Europe/London
tzcode source: internal

attached base packages:
[1] stats     graphics  grDevices utils     datasets  methods   base     

other attached packages:
 [1] sjmisc_2.8.9       readr_2.1.4        ggthemes_4.2.4     ggpubr_0.6.0      
 [5] ggplot2_3.4.2      ggeffects_1.2.2    ehahelper_0.3.9999 dplyr_1.1.2       
 [9] coxme_2.2-18.1     survival_3.5-5     car_3.1-2          carData_3.0-5     
[13] bdsmatrix_1.3-6    lmerTest_3.1-3     lme4_1.1-33        Matrix_1.5-4      

loaded via a namespace (and not attached):
 [1] utf8_1.2.3          generics_0.1.3      tidyr_1.3.0         rstatix_0.7.2      
 [5] stringi_1.7.12      lattice_0.21-8      hms_1.1.3           magrittr_2.0.3     
 [9] grid_4.3.0          backports_1.4.1     gridExtra_2.3       purrr_1.0.1        
[13] fansi_1.0.4         scales_1.2.1        eha_2.10.3          numDeriv_2016.8-1.1
[17] abind_1.4-5         cli_3.6.1           rlang_1.1.1         cowplot_1.1.1      
[21] munsell_0.5.0       splines_4.3.0       withr_2.5.0         tools_4.3.0        
[25] tzdb_0.3.0          nloptr_2.0.3        ggsignif_0.6.4      minqa_1.2.5        
[29] colorspace_2.1-0    sjlabelled_1.2.0    boot_1.3-28.1       broom_1.0.4        
[33] vctrs_0.6.2         R6_2.5.1            lifecycle_1.0.3     stringr_1.5.0      
[37] MASS_7.3-60         insight_0.19.1      pkgconfig_2.0.3     pillar_1.9.0       
[41] gtable_0.3.3        glue_1.6.2          Rcpp_1.0.10         xfun_0.39          
[45] tibble_3.2.1        tidyselect_1.2.0    rstudioapi_0.14     knitr_1.42         
[49] nlme_3.1-162        compiler_4.3.0     

Data files

Data file ADD_I is available on the github page.

Morphometric and Island Comparison

Preparing Data

# read the data file "ADD_I"
ADD_I <- read_csv("Tables/ADD_I.csv", col_types = cols(Island = col_factor(levels = c("CN","AR", "CE", "DS", "FR"))))
# create the earliest born individuals on each island
Isfy <- ADD_I %>% select(founderyear,Island) %>% filter(!duplicated(Island))
# create a data frame with means and sd of each morphometric
ADD_mean <- ADD_I %>% group_by(Island,BirthYear) %>% summarise(RTsd = sd(RightTarsus,na.rm=TRUE),WLsd = sd(WingLength,na.rm=TRUE),BMsd = sd(BodyMass,na.rm=TRUE),HBsd=sd(HeadBill,na.rm=TRUE),RightTarsus=mean(RightTarsus,na.rm=TRUE),WingLength=mean(WingLength,na.rm=TRUE),BodyMass=mean(BodyMass,na.rm=TRUE),HeadBill=mean(HeadBill,na.rm=TRUE))

Linear Mixed Effect Models

Tarsus Length

####RightTarsus ----
IRTlmer <- lmer(RightTarsus ~  Island*fYear + Sex + (1|Observer) + (1|Ageclass) + (1|BirdID), data = ADD_I)
summary(IRTlmer)

#predicting the model 
IRTlmerdata <- ggpredict(IRTlmer,terms=c("fYear [all]","Island","Sex")) %>% rename(fYear=x,RightTarsus=predicted,Island=group,Sex=facet)
IRTlmerdata2 <- merge(IRTlmerdata,Isfy,by="Island", all.x=TRUE) %>% filter(!(Island== "AR" & fYear > 25)) %>% filter(!(Island== "CE" & fYear > 25)) %>% filter(!(Island== "DS" & fYear > 18)) %>% filter(!(Island== "FR" & fYear > 11))%>% mutate(BirthYear = fYear + founderyear) %>% group_by(Island,BirthYear) %>% summarise(RightTarsus = mean(RightTarsus, na.rm=TRUE), std.error = mean(std.error, na.rm=TRUE))

#plotting the model (shown as ggarrange below)
IRTmod <- ggplot(IRTlmerdata2, aes(x = BirthYear, y = RightTarsus, fill = Island)) + 
  geom_point(data = ADD_mean, mapping=aes(x=BirthYear,y=RightTarsus,colour=Island), size = 2) +
  geom_errorbar(data=ADD_mean,aes(ymin=RightTarsus-RTsd,ymax=RightTarsus+RTsd, colour=Island),alpha=0.7) +
  geom_smooth(aes(colour = Island),method = "loess", se = FALSE) + 
  geom_ribbon(aes(ymin=RightTarsus-std.error,ymax=RightTarsus+std.error),alpha=0.5) +
  labs(x = "Birth Year", y= "Tarsus Length (mm)") +
  xlim(1990,2022) +
  theme_tufte(base_size = 15, base_family = "Arial") + 
  scale_color_colorblind() + scale_fill_colorblind()

Body Mass

#### Body Mass ----
IBMlmer <- lmer(BodyMass ~  Island*fYear + Sex  + (1|Observer) + (1|Ageclass) + (1|BirdID), data = ADD_I)
summary(IBMlmer)

#predicting the model
IBMlmerdata <- ggpredict(IBMlmer,terms=c("fYear [all]","Island","Sex")) %>% rename(fYear=x,BodyMass=predicted,Island=group,Sex=facet)
IBMlmerdata2 <- merge(IBMlmerdata,Isfy,by="Island", all.x=TRUE) %>% filter(!(Island== "AR" & fYear > 25)) %>% filter(!(Island== "CE" & fYear > 25)) %>% filter(!(Island== "DS" & fYear > 18)) %>% filter(!(Island== "FR" & fYear > 11))%>% mutate(BirthYear = fYear + founderyear)  %>% group_by(Island,BirthYear) %>% summarise(BodyMass = mean(BodyMass, na.rm=TRUE), std.error = mean(std.error, na.rm=TRUE))

#plotting the model
IBMmod <- ggplot(IBMlmerdata2, aes(x = BirthYear, y = BodyMass, fill = Island)) + 
  geom_point(data = ADD_mean, mapping=aes(x=BirthYear,y=BodyMass,colour=Island), size = 2) +
  geom_errorbar(data=ADD_mean,aes(ymin=BodyMass-BMsd,ymax=BodyMass+BMsd, colour=Island),alpha=0.7) +
  geom_smooth(aes(colour = Island),method = "lm", se = FALSE) + 
  geom_ribbon(aes(ymin=BodyMass-std.error,ymax=BodyMass+std.error),alpha=0.5) +
  labs(x = "Birth Year", y= "Body Mass (g)") +
  xlim(1990,2022) +
  theme_tufte(base_size = 15, base_family = "Arial") + 
  scale_color_colorblind() + scale_fill_colorblind()

Wing Length

#### Wing Length ----
IWLlmer <- lmer(WingLength ~  Island*fYear + Sex + (1|Observer) + (1|Ageclass) + (1|BirdID), data = ADD_I)
summary(IWLlmer)

#predicting the model
IWLlmerdata <- ggpredict(IWLlmer,terms=c("fYear [all]","Island","Sex")) %>% rename(fYear=x,WingLength=predicted,Island=group,Sex=facet)
IWLlmerdata2 <- merge(IWLlmerdata,Isfy,by="Island", all.x=TRUE) %>% filter(!(Island== "AR" & fYear > 25)) %>% filter(!(Island== "CE" & fYear > 25)) %>% filter(!(Island== "DS" & fYear > 18)) %>% filter(!(Island== "FR" & fYear > 11)) %>% mutate(BirthYear = fYear + founderyear) %>% group_by(Island,BirthYear)%>% summarise(WingLength = mean(WingLength, na.rm=TRUE), std.error = mean(std.error, na.rm=TRUE))

#plotting the model
IWLmod <- ggplot(IWLlmerdata2, aes(x = BirthYear, y = WingLength, fill = Island)) + 
  geom_point(data = ADD_mean, mapping=aes(x=BirthYear,y=WingLength,colour=Island), size = 2) +
  geom_errorbar(data=ADD_mean,aes(ymin=WingLength-WLsd,ymax=WingLength+WLsd, colour=Island),alpha=0.7) +
  geom_smooth(aes(colour = Island),method = "lm", se = FALSE) + 
  geom_ribbon(aes(ymin=WingLength-std.error,ymax=WingLength+std.error),alpha=0.5) +
  labs(x = "Birth Year", y= "Wing Length (mm)") +
  xlim(1990,2022) +
  theme_tufte(base_size = 15, base_family = "Arial") + 
  scale_color_colorblind() + scale_fill_colorblind()

Head + Bill Length

#### Head + Bill ----
IHBlmer <- lmer(HeadBill ~  Island*fYear + Sex  + (1|Observer) + (1|Ageclass) + (1|BirdID), data = ADD_I)
summary(IHBlmer)

#predicting the model
IHBlmerdata <- ggpredict(IHBlmer,terms=c("fYear [all]","Island","Sex")) %>% rename(fYear=x,HeadBill=predicted,Island=group,SexEstimate=facet)
IHBlmerdata2 <- merge(IHBlmerdata,Isfy,by="Island", all.x=TRUE) %>% filter(!(Island== "AR" & fYear > 22)) %>% filter(!(Island== "CE" & fYear > 25)) %>% filter(!(Island== "DS" & fYear > 18)) %>% filter(!(Island== "FR" & fYear > 11))%>% mutate(BirthYear = fYear + founderyear)%>% group_by(Island,BirthYear)%>% summarise(HeadBill = mean(HeadBill, na.rm=TRUE), std.error = mean(std.error, na.rm=TRUE))

#plotting the model
IHBmod <- ggplot(IHBlmerdata2, aes(x = BirthYear, y = HeadBill, fill = Island)) + 
  geom_point(data = ADD_mean, mapping=aes(x=BirthYear,y=HeadBill,colour=Island), size = 2) +
  geom_errorbar(data=ADD_mean,aes(ymin=HeadBill-HBsd,ymax=HeadBill+HBsd, colour=Island),alpha=0.7) +
  geom_smooth(aes(colour = Island),method = "lm", se = FALSE) + 
  geom_ribbon(aes(ymin=HeadBill-std.error,ymax=HeadBill+std.error),alpha=0.25) +
  labs(x = "Birth Year", y= "Head + Bill Length (mm)") +
  xlim(1990,2022) +
  theme_tufte(base_size = 15, base_family = "Arial") + 
  scale_color_colorblind() + scale_fill_colorblind()

Arrange and plot

Imods <- ggarrange(IRTmod,IBMmod,IWLmod,IHBmod, common.legend=TRUE, labels = c("A","B","C","D"))
Imods

Lifespan and Reproduction

Preparing data

# make a non-redundant data set for lifespan and reproduction
ADD_Real <- ADD_I %>% filter(Island == "CN") %>% group_by(BirdID) %>% mutate(across(c(RightTarsus,BodyMass,WingLength,HeadBill),mean, na.rm=TRUE)) %>% filter(!duplicated(BirdID)) %>% select(RightTarsus,BodyMass,WingLength,HeadBill,BirdID,SexEstimate,BirthYear,Lifespan,SurvStatus,ReproductiveOutput) %>% mutate(SexEstimate = as.factor(SexEstimate))
# filter for missing values
ADD_tidy <- ADD_Real %>% filter(!is.na(RightTarsus) & !is.na(BodyMass)  & !is.na(WingLength)  & !is.na(HeadBill) & !is.na(SexEstimate)  & !is.na(Lifespan) ) %>% mutate(BodyCondition = BodyMass/RightTarsus) %>% mutate(SexEstimate = as.factor(SexEstimate))

Survival analysis with coxme

Model comparisons

###Lifespan analysis with coxme----

lscox0 <- coxme(Surv(Lifespan,SurvStatus) ~ RightTarsus+ BodyMass + WingLength + HeadBill +SexEstimate + (1|BirthYear), data= ADD_tidy)
lscox1 <- coxme(Surv(Lifespan,SurvStatus) ~ RightTarsus+ poly(BodyMass,2) + WingLength + HeadBill +SexEstimate + (1|BirthYear), data= ADD_tidy)
lscox2 <- coxme(Surv(Lifespan,SurvStatus) ~ RightTarsus*poly(BodyMass,2) + WingLength + HeadBill +SexEstimate + (1|BirthYear), data= ADD_tidy)
lscox3 <- coxme(Surv(Lifespan,SurvStatus) ~ poly(BodyCondition,2) + WingLength + HeadBill +SexEstimate + (1|BirthYear), data= ADD_tidy)
lscox4 <- coxme(Surv(Lifespan,SurvStatus) ~ RightTarsus+ poly(BodyMass,2)*SexEstimate + WingLength + HeadBill + (1|BirthYear), data= ADD_tidy)

AIC(lscox0,lscox1,lscox2,lscox3,lscox4) # lscox4 best
BIC(lscox0,lscox1,lscox2,lscox3,lscox4) # lscox4 best
# lscox4 was selected based on the best AIC and BIC values

lscox4

Predicting and plotting coxme

####predict lifespan----
ADD_round <- ADD_Real %>%mutate(across(c(RightTarsus, BodyMass, HeadBill, WingLength), round, 0)) %>% distinct(RightTarsus, BodyMass, HeadBill, WingLength,SexEstimate) %>% filter(!is.na(RightTarsus) & !is.na(BodyMass)  & !is.na(WingLength)  & !is.na(HeadBill) & !is.na(SexEstimate))
ADD_end <- bind_rows(ADD_tidy,ADD_round)
rs <- predict_coxme(lscox4,newdata = ADD_end,type="risk", se.fit=TRUE)
rsend <- ADD_end %>% add_columns(rr = rs$fit[,1], se.fit = rs$se.fit[,1])
ADD_tail <- rsend %>% tail(nrow(ADD_round)) %>% mutate(SexEstimate = as.factor(SexEstimate)) %>% group_by(BodyMass,SexEstimate) %>% summarise(mean_rr=mean(rr),mean_se.fit=mean(se.fit))

# plotting coxme
coxmeplot <- ggplot(ADD_tail,aes(x=BodyMass,y=mean_rr)) + 
  geom_line(aes(colour=SexEstimate)) +
  geom_ribbon(aes(fill=SexEstimate,ymin=mean_rr-mean_se.fit,ymax=mean_rr+mean_se.fit), alpha = 0.25) + 
  labs(x = "Body Mass (g)", y = "Mortality Risk Score") +
  guides(fill=guide_legend(title="Sex"), colour=guide_legend(title="Sex")) +
  theme_tufte(base_size = 15, base_family = "Arial") + 
  scale_color_colorblind(labels = c("Female", "Male")) +
  scale_fill_colorblind(labels = c("Female", "Male"))

Reproductive Output using GLM

Model comparisons

###Reproduction glm----
ROglmF <- glm(ReproductiveOutput ~ RightTarsus+ BodyMass + WingLength + HeadBill  ,data = ADD_tidy, family = "poisson")
ROglmR <- glm(ReproductiveOutput ~ poly(Lifespan,2) + SexEstimate + BirthYear,data = ADD_tidy, family = "poisson")
ROglm0 <- glm(ReproductiveOutput ~ RightTarsus+ BodyMass + WingLength + HeadBill  + Lifespan + SexEstimate + BirthYear,data = ADD_tidy, family = "poisson")
ROglm1 <- glm(ReproductiveOutput ~ RightTarsus+ BodyMass + WingLength  + HeadBill  + poly(Lifespan,2) + SexEstimate + BirthYear,data = ADD_tidy, family = "poisson")
ROglm2 <- glm(ReproductiveOutput ~ poly(RightTarsus,2)+ BodyMass + WingLength + HeadBill  + poly(Lifespan,2) + SexEstimate + BirthYear,data = ADD_tidy, family = "poisson")
ROglm3 <- glm(ReproductiveOutput ~ RightTarsus+ poly(BodyMass,2) + WingLength + HeadBill  + poly(Lifespan,2) + SexEstimate + BirthYear,data = ADD_tidy, family = "poisson")
ROglm4 <- glm(ReproductiveOutput ~ RightTarsus+ BodyMass + poly(WingLength,2) + HeadBill  + poly(Lifespan,2) + SexEstimate + BirthYear,data = ADD_tidy, family = "poisson")
ROglm5 <- glm(ReproductiveOutput ~ RightTarsus+ BodyMass + WingLength + poly(HeadBill,2)  + poly(Lifespan,2) + SexEstimate + BirthYear,data = ADD_tidy, family = "poisson")

AIC(ROglmF,ROglmR,ROglm0,ROglm1,ROglm2,ROglm3,ROglm4,ROglm5) # ROglm4 slightly better than ROglm1 second
BIC(ROglmF,ROglmR,ROglm0,ROglm1,ROglm2,ROglm3,ROglm4,ROglm5) # ROglm1 best 
#ROglm1 was selected based on AIC and BIC scores
vif(ROglm1)

summary(ROglm1)

Predicting and plotting GLM

####predict reproduction----
# grouping variables for glm plot
ROplotdata <- ADD_tidy %>% mutate(BodyMass = case_when(BodyMass > 16.5 ~ "16.5", BodyMass < 14.5 ~ "14.5",BodyMass < 16.5 & BodyMass > 14.5 ~ "15.5")) %>% filter(!is.na(BodyMass))

# predicting glm 
ROglmdat <- ggpredict(ROglm1, terms = c("Lifespan","BodyMass")) %>% rename(Lifespan=x,ReproductiveOutput=predicted,BodyMass=group)

#plotting glm
ROglmplot <- ggplot(ROglmdat, aes(x=Lifespan,y=ReproductiveOutput)) + 
  geom_point(data = ROplotdata, mapping=aes(x=Lifespan,y=ReproductiveOutput,colour=BodyMass),position="jitter") +
  geom_smooth(aes(colour=BodyMass), se = FALSE) + 
  geom_ribbon(aes(ymin=conf.low,ymax=conf.high,fill=BodyMass), alpha = 0.25) +
  labs(x = "Lifespan (years)", y= "Total Number of Offspring") +
  theme_tufte(base_size = 15, base_family = "Arial") +
  scale_color_colorblind() + scale_fill_colorblind()

Arrange and plot

fitplot <- ggarrange(coxmeplot,ROglmplot,  labels = c("A","B"))
fitplot

Time taken to run the whole script

Acknowledgment

Thank you for going through the code and supporting open research. This paper would not have been possible without all authors contributing towards it. Massive thanks to everyone that has shared their excitement for this paper.